Observability vendors race to make AI operational

AI shifted from narrative to packaging

The fortnight was not quiet. It was a coordinated push to make AI feel like an operational feature rather than a slideware theme. On 28 Sep, PagerDuty introduced an 'AI Operations Process Map'. On 25 Sep, New Relic published new observability and AI posts, including a guide for building SRE agents and a 2026 forecast, after adding a page on the general availability of New Relic Lens on 21 Sep. On 22 Sep, Sentry launched a product page for an AI integration for Laravel, then on 26 Sep posted three San Francisco engineering roles whose descriptions emphasised an AI-native future.

The hiring trail points in the same direction. Datadog listed seven new roles on 20 Sep, including five engineering roles related to machine learning, then on 27 Sep listed 13 more openings, including seven engineering roles tied to machine learning and AI. Grafana Labs listed seven remote Engineering Manager roles in Observability on 22 Sep, 20 engineering and sales openings on 25 Sep, and three AI/ML-focused roles on 27 Sep. This is not a single launch cycle. It is category-wide budget allocation.

The important distinction is where AI appeared. It showed up in incident management content at PagerDuty, SRE agent content at New Relic, developer workflow pages at Sentry, AI/ML hiring at Datadog and Grafana Labs, and Grafana videos on agents, OpenTelemetry and AI versus manual dashboard creation on 19 Sep. The vendors are converging on the same promise: fewer humans staring at noisy systems, more automated judgement around production.

The ad market looked fully loaded

On 26 Sep, all five vendors showed 40 Google ad creatives running. The refresh was broad: Datadog had 37 new creatives, New Relic 37, Sentry 29, Grafana Labs 23 and PagerDuty 18. On 28 Sep, all five still showed 40 creatives running, with Datadog adding 31 new creatives whose earliest first shown date was 21 Sep, New Relic 19, Sentry 13, PagerDuty eight and Grafana Labs six.

That simultaneity matters. The category is not waiting for one flagship conference or one end-of-quarter campaign. Distribution is being continuously tuned while the product surfaces change underneath it. On 25 Sep, Sentry changed its website call-to-action from 'Integrations' to 'Tracing'. On 27 Sep, New Relic added an 'Instant Observability' product page for PostgreSQL. On 26 Sep, Datadog published five blog posts across production alerts, agent evaluation, product analytics for Shopify and Salesforce, remote configuration for RUM, and log parsing.

The quiet signal is not any one page. It is the speed of rotation. A buyer comparing these tools in late September would have seen changing homepages, new product pages, fresh technical content and nearly full ad shelves at the same time. That is what a crowded category looks like when differentiation has to be manufactured weekly.

Packaging became the other battleground

PagerDuty made the most visible pricing moves. On 23 Sep, it launched a blog page on automating incident management with PagerDuty and Slack and removed its pricing page from the website. On 26 Sep, it changed the Free tier by adding five users, a basic chat experience, limited email, push, SMS and phone notifications, up to five users per month, 100 monthly international phone and SMS notifications, API calls included and more than 750 out-of-box integrations, while removing 'No maintenance windows ever' from the Free tier. On 28 Sep, it added a new pricing page and removed four existing pricing pages tied to AIOps, customer service operations, incident management and process automation.

Grafana Labs also adjusted packaging language on 19 Sep, adding 'Everything in Free plus' to the Pro tier. The change is small in isolation, but it sits next to a larger commercial push: on the same date Grafana Labs listed 29 Commercial and Enterprise Account Executive openings across remote locations including Israel, Italy, Spain and the UK, then added further engineering and sales postings during the following week.

Pricing-page churn is often more revealing than pricing itself. PagerDuty did not simply alter a feature table. It removed, reintroduced and consolidated pages across the exact period in which it was also publishing AI operations content and appointing a new Vice President of EMEA on 26 Sep. The public surface suggests a company tightening how it presents the product family, not just changing a SKU.

Reliability vendors had reliability events

The category also supplied its own reminder of the stakes. Sentry reported two incidents on 19 Sep: a critical EU error ingestion delay lasting about 200 minutes and a major spans ingestion delay lasting about 41 minutes. On 24 Sep, CVE-2026-83803 was published for Sentry versions 23.11.0 to 26.7.0, covering unsafe deserialisation of a legacy database when the relocation feature is enabled.

Grafana Labs reported a minor GCP US Central cloud provider integrations latency incident on 19 Sep, then on 25 Sep reported two resolved incidents: incorrect billing and usage for Kubernetes observability lasting about 999 minutes, and a major access issue for a small group of users lasting about 1,210 minutes. On 27 Sep, it reported a minor hosted Azure metrics issue affecting Cloud Provider Observability that was resolved after about 49 minutes. Grafana Labs also released version 13.2.2 on 25 Sep with security fixes for CVE-2026-15815, CVE-2026-76154 and CVE-2026-79656.

New Relic reported a no-impact Salesforce service disruption on 18 Sep lasting about 332 minutes, a one-minute test incident on 24 Sep, and on 27 Sep a major US-region data irregularities incident lasting about 156 minutes plus a minor UI errors incident that was resolved almost immediately. Datadog reported minor incidents on 20 Sep for delayed email-triggered events lasting about 96 minutes, on 23 Sep for delayed monitor notifications lasting about 46 minutes, and on 26 Sep for users being unable to acknowledge, escalate or resolve On-Call Pages for about 27 minutes. On 19 Sep, two high-severity Datadog CVEs were reported, CVE-2026-50275 in PHP Tracer and CVE-2026-50277 in dd-trace-cpp, both with CVSS 7.5.

This is not a scandal. It is the operating reality of companies selling production truth while running large, complex production systems themselves. The market reads these events less as isolated outages than as proof points for response discipline, disclosure habits and the quality of the recovery loop.

Open source stayed central to the funnel

Grafana Labs continued to turn GitHub into product distribution. On 21 Sep, total stars across its top repositories were reported up 2,876 percent to 98,979, alongside new repositories including ai-sdk, grafana, gcx, alloy, tempo, pyroscope and grafanacon2026-related work. On 25 Sep, total stars across top repositories reached 141,524 after an increase of 42,545, and version 13.2.2 shipped with security fixes and a dashboard query-variable bug fix. On 27 Sep, Grafana Labs launched further repositories including lezer-promql-tree-viz, yugabyte-datasource, schemads, tempo, x-ray-datasource, opentelemetry-collector-extras, mimir-prometheus, tanka and sustainability-o11y.

Sentry was similarly active. On 19 Sep, it released v0.1.0 of action-filter-paths and introduced repositories including action-filter-paths, sentry-native, sentry-dotnet, symx, outpost and praise. On 22 Sep, it released version 26.9.0 and introduced repositories including sentry-android-gradle-plugin, sentry-python, sentry-dart and sentry-fastlane-plugin. On 26 Sep, stars across top repositories rose from 53,546 to 65,539, an increase of 11,993, and on 28 Sep it released version 0.4.2 of coverage-action with hardened gitignore exclusions and a coverage file browser.

New Relic also pushed developer surfaces. On 19 Sep, it released version 9.4.0 of the Java agent. On 21 Sep, it released version 13.6.0 of the Python agent with support for OpenSearch, Microsoft Azure SQL, new Kinesis methods and a new browser_monitoring.version setting. On 27 Sep, it released version 3.99.3 of terraform-provider-newrelic with documentation updates for audit log ingestion and cloud integrations. In observability, open source is still not peripheral. It is where credibility, adoption and technical taste are all negotiated.